IP3 2022 – Video Processing 2 (LOT 13765)
This lot is generally related to a digital image processing system for identification of multi-class objects using boosted classifiers. Disclosed is a training system for identification of multi-class objects in an image or a video sequence by determining an ensemble of weak binary classifiers which are common to a set of object classes so that each object class shares a common set of features with the other object classes. Each feature of the common set features is associated with a weak binary classifier. Also disclosed are techniques to form individual strong classifiers by adapting a separate set of weights for the ensemble of weak classifiers. The technology may be implemented in video processing systems, real-time vehicle detection and tracking systems, etc.